The variants selected in calculating PGSAUD were based on P values, effect sizes, and linkage disequilibrium from two meta-analyses, after filtering for the same direction of effect. No functional annotations were used due to the limited knowledge of genetic mechanisms contributing to risk for AUD. Furthermore, the performance of PGSAUD depends on the statistical power of the discovery datasets, which are still small and therefore the PGSAUD is an imprecise estimate of genetic liability for AUD. Additionally, it is worth noting that within-cohort variability in the AUD GWAS in MVP-AUD and UKBB-AUDIT-P that may not have accounted for all possible confounders may have influenced effect size estimation in those cohorts. In Both GWAS, sensitivity analyses were conducted to provide assurances that this was not the case. However, it is plausible that any residual effect estimation bias in the individual GWAS may have impacted our PGSAUD analyses in COGA and Indiana Biobank, if, for instance, our regression missed a specifically influential confounder. Environmental influences and psychiatric comorbidities, as well as their interactions with genetic factors are also significant contributors to AUD.
